Cole Wyeth's Shortform

Cole Wyeth

This is a special post for quick takes by Cole Wyeth. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.

My preference ordering over approaches to surviving the next decade, assuming I had unlimited political capital, is (don't build ASI and):

1: Solve mind uploading

2: ... or at least intelligence amplification

(before proceeding, and then maybe do)

3: Imitation learning (with radically improved theoretical foundations)

4: ... or corrigibility / oracles

(and don't)

5: Try to build aligned (or controlled) autonomous ASI agents

Unfortunately 5 seems to be profitable, in the short term, so I guess that's what we're doing. In fact, the "plan" seems to be the exact reverse or so of my preference ordering on what it should be.

3dr_s11mo

I think it's just a matter of what's more technologically achievable. Building LLMs turned out to be a lot easier than understanding neuroscience to a level even remotely close to what's necessary to achieve 1 or 2. And both of those also require huge political capital due to needing (likely dangerous) human experimentation that would currently be considered unacceptable.

2dr_s11mo

It would be slower for sure, at least, being bound to human dynamics. But "same problems but slower" isn't the same as a solution/alternative. Admittedly better in the limited sense that it's less likely to end with straight up extinction, but it's a rather grim world either way.

Since it's slower, the tech development cycle is faster in comparison. Tech development --> less expensive tech --> more access --> less concetration of power --> more moral outcomes.